Chat The harness How it works Stack Can and can't Platforms Models Pricing Company Careers About Request trial

By Raman, Jayson, Ashly, and Emily

Private trials Public beta Q4 2027
September 2026 Models you already run, plus Apple

Point the harness at Grok, Gemini, OpenAI, Claude, Qwen, DeepSeek, Llama, Mistral, Ollama, and any OpenAI-compatible local server. Same install on macOS, iOS, and watchOS. Glance a task, approve a write from the wrist.

Grok Gemini OpenAI Claude Ollama Llama Mistral Qwen DeepSeek

An agent harness that runs
on the machine you already use.

computer.vodka is the control loop around the model you already run. We are a young company building the next layer of local AI: a harness that gives every turn the same disciplined flow, plan, choose one tool, read the result, then either stop or ask for approval before a hard-to-reverse action. Tools, memory, and approvals live in one folder on your disk. Install it on Windows, Linux, Android, and the Apple ecosystem. The install is free. Hosted extras are optional and metered.

Public beta opens Q4 2027, after private-trial bugs are actually fixed. Thanks for waiting.

Request the trial What it can and can't do
Windows Apple Linux Android

Same harness on Windows, Linux, Android, and the Apple ecosystem, macOS, iOS, and watchOS.

computer.vodka · session probe
The harness, in a chat

You type the task. The harness runs the loop.

A harness is the runtime around a model: planner, tool runner, workspace root, session log, and a permission gate. You talk here. It proposes a plan, calls one tool, reads the output, and waits if the next step would delete, overwrite, install, or send. Quality follows the weights you point it at. This page is a preview of that chat shell. The real loop starts after you install.

computer.vodka
Online
Supported apps

These are the connectors the harness can switch on inside the chat. Live OAuth ships with the trial. Hosted extras keep them alive when the laptop is closed.

The harness

Not a solo agent. The loop around any model you run.

Most products hide the model and ship your files to a GPU farm. A harness does the opposite. It is a small process on your machine: a planner, a strict tool schema, markdown memory, and a gate in front of writes. You attach Llama, Qwen, Mistral, Gemma, or an OpenAI-compatible server on localhost or your LAN. Swap the weights and the product does not change. Free local quality is exactly as good as those weights. The subscription only covers work a closed laptop cannot do alone, like always-on text and multi-device sync.

A model-agnostic agent loop

Plan, pick one tool, observe, repeat. Talks to Ollama, llama.cpp, LM Studio, vLLM, or any OpenAI-compatible /v1/chat/completions endpoint on localhost or your LAN. Swap the weights without rewriting the product.

Your files stay put

Every path is resolved against one workspace folder. Memory is markdown you can open. The installed harness does not phone home with prompts or file contents.

Trimmed for a laptop

Rolling summaries, a small tool schema, and interruptible streams so a long session still fits on Apple Silicon or a strong CPU.

Approval before damage

Deletes, overwrites, installs, and outbound sends wait for a yes. You see the plan and the diff before anything hits disk.

Tools, not a plugin zoo

read, write, search, shell, fetch, diff, plus connectors you toggle in the chat. Enough to finish work. Not a kitchen sink.

Apps as switches

Mail, calendar, docs, chat, and the rest sit behind the Apps toggle. Off means the tool is never offered to the model.

Text the harness

Give it the number you want watched. SMS and WhatsApp then drop a task into the same loop. The first message in a thread is password-checked so a random sender cannot act for you.

Free local, hosted extras on a sub

The loop on your disk is free and follows your model. Cloud fallback, multi-device sync, always-on text, and scheduled jobs are usage-billed.

How it works

Three steps, then the harness lives on your desk.

No account to babysit for the local path. You go from a trial request to a working loop in one folder. After that, every turn is plan, one tool call, observe, then continue or stop. Quality is the model you attach. Hosted extras stay optional.

1

Ask the chat for a trial

Open the interface below and tell the agent you want in, or just type your email. It collects what it needs and sends the request for you.

2

Install it on your devices

Install on Windows, Linux, Android, and Apple (macOS, iOS, watchOS). Point it at Ollama, llama.cpp, LM Studio, or any OpenAI-compatible endpoint. SMS, WhatsApp, and optional sync use the same loop. After setup the local loop does not need an account.

3

Type the task

Point it at a folder, say what you want in plain language, and let the loop plan. Approve writes, deletes, and installs. Read the transcript after.

How the harness is built

A local runtime, split into parts you can inspect.

The site is a preview. The installed product is one process: planner, tool runner, memory on disk, policy gate. That is the whole idea of a harness. It does not pretend to be the model. It wraps whatever you attach, keeps paths inside one workspace root, and writes a plain transcript of every tool call. Free local quality tracks those weights. Subscription extras are only the hosted pieces a single disk cannot cover.

The loop

Every turn is the same four steps: plan, pick one tool, read the result, then decide whether to continue or stop. Context is trimmed each turn so a long session still fits on a laptop GPU or a strong CPU.

  • 01Talks to Ollama, llama.cpp, LM Studio, or vLLM over any OpenAI-style /v1/chat/completions endpoint on localhost or your LAN.
  • 02Model-agnostic: Llama, Qwen, Mistral, or Gemma, and whatever you already run. Quality follows the weights, not a hidden API.
  • 03Tool calls use strict JSON schemas. A malformed call is caught and retried instead of crashing the turn.
  • 04Context window is managed with a rolling summary, so multi-step tasks stay in one transcript without overflowing the model.
  • 05Streaming tokens render live, and any turn can be interrupted mid-run without corrupting the workspace.
  • 05bOptional hosted router on a subscription: if the local model stalls on a hard turn, that one step can bounce to a cloud model. Files stay in the workspace unless you attach them.
  • 05cSystem prompt and tool list are generated from the capability switches, so a disabled app is invisible to the model, not just ignored.

Tools

Enough to finish a job in a folder, not a kitchen-sink plugin list. Risky calls wait for a yes.

readwritesearchshellfetchdiffconnectmailcalendarsms
  • 06Workspace is a single folder. Every path is resolved and checked against that root, so it cannot walk your home directory.
  • 07Writes go through a diff preview. Deletes, overwrites, and install commands pause until you confirm the plan.
  • 08Shell runs with a timeout and captured stdout/stderr, so a hung command can't lock the session.
  • 08bFetch is allow-listed to URLs you ask for. It is a page reader, not a crawler. It does not wander the open web on its own.
  • 08cConnectors speak scoped OAuth. Tokens live in a local key store. Revoke per app without touching the rest of the harness.

Memory

No hidden vector store you cannot open. Notes live as plain markdown in the same workspace.

  • 09Memory is a set of .md files you can open, edit, or delete like any other file.
  • 10Optional local embeddings index the workspace for search. It stays on disk and never leaves the machine.
  • 11Every tool call is appended to a plain-text session log, so you can audit exactly what ran, in order.
  • 11bNotes can be pinned into the next turn as working memory. Nothing is buried in a vendor vector store you cannot open.

What stays local

Prompts, file contents, shell output, and embeddings stay on the box you run. This page only emails a trial request. The installed agent does not phone home.

  • 12Install on Windows, Linux, Android, and the Apple ecosystem (macOS, iOS, watchOS). Same harness, same loop. The site is built for those screens too.
  • 13No telemetry and no account after setup. You can run it fully air-gapped once the model is pulled.
  • 14Config is one readable file. No cloud project, no API key required for the local path.
  • 14bFirst-class on Apple Silicon, which is what we develop on. CUDA and CPU backends through the model server you already chose.

Connectors and text-in

Apps in the chat are switches, not a second control panel. Live connections ship with the trial. You can also text the agent once a conversation is unlocked.

  • 15Mail, calendar, docs, and chat apps sit behind the Apps toggle. Off means the tool is never even offered to the model.
  • 16Connectors use scoped OAuth. Tokens are stored locally and you revoke access per app whenever you want.
  • 17The first message in any text thread is password-checked, so a random number cannot act on your behalf.
Platforms

Install it on every device you already use.

The harness installs on Windows, Linux, Android, and the Apple ecosystem. Same loop on each: plan, one tool, observe, stop. Desktop gets the full folder manager. Phones get the same chat, inbound SMS or WhatsApp, and optional sync. On Apple, iOS and watchOS share that loop so you can glance status and approve a step from the wrist.

Desktop

Windows

Install the local manager, point it at Ollama, LM Studio, or any OpenAI-compatible server on the box or LAN. Full read, write, shell, and approval gate on one workspace folder.

Apple ecosystem

macOS, iOS, watchOS

One Apple path. First-class on Apple Silicon. Mac runs the full folder manager. iOS keeps the same chat, inbound number, and approvals. watchOS shows the current task and lets you confirm a write or send without opening the phone. Intel Macs use the same binary.

Desktop

Linux

Same local manager, same folder root. Works next to llama.cpp, vLLM, or Ollama on a workstation or a box on the LAN. No account after setup. Air-gap is fine once the model is pulled.

Phone

Android

Install the harness on Android. Use the same chat, or text it from WhatsApp or SMS after a password check. Hosted extras keep the loop alive when the laptop lid is closed.

Local manager

One loop, every screen

Plan, one tool, observe, stop. That loop does not change. Desktop runs it on disk. Phones run the same manager. Quality still follows the model you pointed at, not a hidden farm.

Models

Attach the weights you already use.

The harness is not a branded brain. It talks to Grok, Gemini, OpenAI, Claude, and the local and open-source stack you already run: Ollama, llama.cpp, LM Studio, Llama, Qwen, DeepSeek, Mistral, Gemma. Hosted extras can bounce a hard turn to a cloud model. Files stay in the workspace unless you attach them.

Grok · xAI

OpenAI-compatible endpoint. Point the harness at the API you already have.

Gemini

Google models through a compatible server or hosted fallback on a hard turn.

OpenAI

Any /v1/chat/completions server, including official and local proxies.

Claude

Anthropic through a compatible gateway. Same plan, tool, observe loop.

Ollama

Local first. Pull Llama, Qwen, Mistral, Gemma on the box and attach.

Llama

Meta weights through Ollama, llama.cpp, LM Studio, or vLLM.

Mistral

Open weights on your machine. Quality follows what you load.

Qwen

Alibaba open models. Strong local option on Apple Silicon and CUDA.

DeepSeek

Open-source reasoning models you can run locally or on the LAN.

LM Studio

Desktop server. Same OpenAI-style endpoint the harness already speaks.

Hugging Face

Weights you already downloaded. Serve them with llama.cpp or vLLM.

Gemma

Google open weights on the local path. Swap them without changing the product.

Being straight with you

What the harness can do, and what it cannot yet.

Type a task and the loop tries it: plan, tool, observe, stop. How good the answer is depends on the model you attached, because the harness is not a secret smarter model. Free local follows those weights. The subscription adds hosted reasoning, sync, and always-on text. Video generation and native video edit are the clear not-yet. Almost everything else is already in the runtime.

It can

  • Read and edit a workspace folder

    List, search, open, rewrite, and grep across one directory you choose. Paths cannot walk outside that root.

  • Write drafts, data, and scripts

    Markdown, JSON, CSV, HTML, Python, R, shell, SQL. Diff preview before anything stays on disk.

  • Run local commands you approve

    Shell, tests, formatters, builds, short programs. Installs and deletes wait for a yes.

  • Wrap any model you already run

    Ollama, llama.cpp, LM Studio, vLLM, or any OpenAI-compatible server. The harness is the same. The brain is yours.

  • Keep a readable session log

    Every tool call lands in a local transcript. Open it like any other file. Audit what ran, in order.

  • Keep working offline after setup

    Once the binary and model are on the machine, prompts and files do not go to our servers.

  • Run across the machines you already own

    Install on Windows, Linux, Android, and the Apple ecosystem. Same loop, plus SMS, WhatsApp, and optional sync.

  • Talk to connected apps

    Mail, calendar, docs, chat, drive, and the rest of the ticker below the chat. Off means the tool is never offered.

  • Pick up a task from a text

    SMS or WhatsApp, after a password check on the first message. Hosted gateway is a subscription extra so it works when the lid is closed.

  • Research, summarise, and extract

    Open a page you name, pull quotes, turn them into notes. With local files on: grind a folder of PDFs, CSVs, or code inside the workspace root.

  • Code, debug, and test in the folder

    Read a repo, patch a file, run the test command you approve, paste the failure back into the next turn.

  • Draft mail, briefs, and reports

    Write the thing, show it, wait for send. Connected inboxes are a connector, not a default outbound pipe.

  • Images the model already supports

    Describe, caption, and file-handle images in the workspace. Generate stills only if the attached model can. That is the model, not a hidden API.

  • Tables, spreadsheets, and light analysis

    Read CSV or Excel-like files, compute, chart to a file, write a summary. Heavy stats still want a model that can count.

  • Remember across turns on disk

    Plain markdown notes plus an optional local embeddings index. You can delete a memory by deleting the file.

It can't, at least not yet

  • Generate or edit video

    No native video model in the harness yet. It can organise clips on disk and write ffmpeg command plans you approve. The pixels themselves are a later problem.

  • Match frontier cloud models on the free path

    A 7B or 14B on a laptop is not GPT-class reasoning. Subscription hosted extras exist for the hard turns. Files stay local unless you attach them.

  • Run well on a weak machine

    Useful local models need RAM and, ideally, a GPU. Older phones and low-RAM laptops will stall.

  • Browse the open web by default

    Network fetch is off until you turn it on, and even then it is for pages you ask it to open, not a general crawler.

  • Always read the task correctly

    It can mis-scope a folder, skip a file, or write the wrong thing. Review the diff before you approve.

  • Replace git or backups

    There is no built-in version control. Keep your own copies of anything you cannot afford to lose.

  • Act from this website

    The chat on this page is a preview. It cannot see your disk. The loop starts after you install the trial.

  • Be a single branded brain

    computer.vodka is the harness. It is not "the model." If the weights are thin, the answers are thin. That is the deal on the free path.

Pricing

The harness stays free. Hosted extras are usage-only.

Install it, point it at your model, work in a folder. No card. The free path is the full local loop: plan, tool, observe, approve. How well that works is how well your local or LAN model works. From the Q4 2027 public beta, hosted extras are optional and metered. Cloud reasoning for hard turns, multi-device sync, always-on text, live connectors, jobs that outlive sleep. No fixed seat fee. Quiet months stay cheap. Anything that can run on your machine stays in the local product.

Always free

Stays local

No subscription · ships with the trial · quality = your model

  • The full harness loop

    Plan, tool, observe, stop. Read, write, search, shell, and local fetch inside the folder you choose.

  • Your own model

    Ollama, llama.cpp, LM Studio, or any OpenAI-compatible endpoint on the machine or LAN.

  • Memory and logs on disk

    Markdown notes and a plain transcript. You can open or delete them without an account.

  • Approval before damage

    Deletes, overwrites, and installs wait for a yes. That gate does not move to the cloud.

  • Works offline after setup

    No seat, no monthly fee, no phone-home just to keep using the agent on one computer.

Q4 2027 · pay as you go

Hosted extras

Usage billed each month · no fixed fee, no send limit, only what you use

  • Cloud model when local is not enough

    Metered hosted reasoning for the hard turns. Files stay in your workspace unless you attach them.

  • Same session on more than one device

    Sync chat and memory notes across a laptop and a phone. A single disk cannot do that by itself.

  • Always-on text and WhatsApp

    A hosted gateway so the agent can pick up SMS or chat when the laptop is closed. Bill by messages handled, not a plan cap.

  • Live app connectors

    Gmail, Calendar, Slack, Drive, and similar need an OAuth broker we host. You pay for the calls you make.

  • Jobs that outlive sleep

    A hosted runner keeps a long task going after the lid closes. Local processes die with the machine.

  • Team approvals and shared logs

    A second person can approve a risky step and read the same transcript. That is a server product.

  • Usage dashboard and receipts

    See tokens, connector calls, and messages for the month. Export a receipt. Stop anytime and local keeps working.

  • Scheduled jobs and webhooks

    Wake the agent on a timer or an inbound event. Hosted only, because a closed laptop cannot keep a clock.

About us

A young company building the next layer of local AI.

We design, ship, and run the harness on our own machines. Four people, one product: the control loop around the model you already use. We are not a lab wrapping someone else’s API. We are building the runtime people will keep when the hype cycle moves on.

R

Raman

Founder · Runtime and inferenceNew Zealand

Builds the actual loop. Plan, one tool, read the output, stop. He wires Ollama, llama.cpp, LM Studio, and any OpenAI-style /v1 endpoint so a laptop model feels like a product, not a terminal demo. Days go into Apple Silicon trim, interruptible streams, and a tool runner that does not stall a long session. Install paths on Windows, Linux, Android, and Apple are his too. If the probe lies about the box, he wants to know.

J

Jayson

Co-founder · Product and evaluationAustralia

Makes sure a task actually finishes for someone who does not live in a terminal. He writes folder evals, kills demo-only tricks, and holds the Q4 2027 public beta until private-trial bugs are gone. Hosted extras sit with him too: sync, always-on text, cloud fallback when a local model stalls. Those stay metered and optional. The free loop on disk does not get taxed for existing.

A

Ashly

Co-founder · Design and interfaceUnited Kingdom

Owns the shell you tap. Chat, capability switches, approval states, the mark, and this trial flow. She builds the desktop and phone UI so it reads as a tool, not a raw CLI, and the watch glance so you can confirm a write from the wrist. Type, motion, and the one blue accent are hers. No rainbow, no circus. If a control feels loud, it does not ship.

E

Emily

Co-founder · Memory and tool safetyUnited States

Keeps the dangerous bits from running loose. Memory is markdown in the workspace, not a hidden vector store. Every path resolves against one folder root, so a bad plan cannot walk your home directory. Writes, deletes, installs, and outbound sends wait for a yes. She also built the first-message password check on SMS and WhatsApp, so a random number cannot act for you.

Company

A small studio building a local runtime, not a model farm.

computer.vodka is a young company. Four people ship the harness on the machines they already use. We are incorporated to sell a local product: planner, tool runner, workspace root, and an approval gate. Contact is one inbox. We answer it.

What we sell

The install is free. Quality follows the model you attach. Hosted extras from the Q4 2027 public beta are optional and metered: cloud fallback, sync, always-on text, live connectors, jobs that outlive sleep.

How we work

Remote-first across New Zealand, Australia, the United Kingdom, and the United States. We design on Apple Silicon, test on Windows, Linux, Android, iOS, and watchOS, and keep the loop the same on every screen.

Contact

Trials, press, partnerships, and security notes all go to [email protected]. Put the topic in the subject. We do not ask for a form, a Slack, or a calendar link first.

Private trials open Public beta Q4 2027 [email protected] NZ · AU · UK · US
Careers

A few seats. Real work on the loop.

We hire when a gap is blocking the product, not to fill a headcount slide. Roles are remote across NZ, AU, UK, and US time zones. You will ship against a workspace folder, a strict tool schema, and a permission gate. Apply by email. Put the role name in the subject and send it to [email protected].

Runtime engineer

Full-timeRemoteInference

Own the plan / one tool / observe / stop loop. Talk to local and LAN model servers. Trim context, stream tokens, keep a long session alive on Apple Silicon or a strong CPU without stalling the tool runner.

You will

  • Wire Ollama, llama.cpp, LM Studio, vLLM, and any OpenAI-compatible /v1/chat/completions endpoint
  • Manage rolling summaries so multi-step tasks stay inside a laptop context window
  • Catch malformed tool JSON and retry the turn instead of crashing the session
  • Ship interruptible streams on Windows, Linux, Android, and Apple

Skills we need

  • Systems or backend work in Rust, Go, or modern C++
  • Streaming HTTP, JSON schemas, and cancellation that does not corrupt state
  • Enough ML serving to reason about tokens, context, and GPU vs CPU backends
  • Comfort reading traces when a model lies about a tool call

Tools you should already use

Rust or Go or C++Ollamallama.cppLM StudiovLLMOpenAI-compatible APIsGitlldb / gdbApple SiliconCUDA optional
Apply · [email protected]

Tool safety engineer

Full-timeRemotePolicy

Keep writes, deletes, installs, and outbound sends behind a yes. Every path resolves against one workspace root. Memory stays markdown. A random number cannot act for you.

You will

  • Design the approval gate and the diff preview before anything hits disk
  • Sandbox shell with timeouts and captured stdout / stderr
  • Build the first-message password check on SMS and WhatsApp
  • Keep a plain transcript of every tool call a human can open and audit

Skills we need

  • Threat modeling for agents that can write files and run commands
  • Path canonicalization, allow-lists, and least-privilege tool schemas
  • Secure local storage for tokens and session secrets
  • Clear writing: a policy that cannot be explained does not ship

Tools you should already use

Rust or GoOS sandboxingseccomp / AppArmor / sandbox-execJSON SchemaOpenSSL / keychaingit-diffregex + path APIsmarkdown
Apply · [email protected]

Product designer

Full-timeRemoteInterface

Chat shell, capability switches, approval states, desktop and phone, watch glance. One blue accent. If a control feels loud it does not ship. You sit next to the runtime, not above it.

You will

  • Design the installed chat, panels, and approval UI across desktop, iOS, Android, and watchOS
  • Make plan / tool / observe readable in one glance, including a wrist confirm
  • Own type, motion, and the mark. No rainbow, no circus
  • Prototype in the product, not only in a file that never runs

Skills we need

  • Product UI for tools, not marketing pages
  • Information density without clutter: diffs, logs, capability chips
  • Accessible tap targets, reduced-motion paths, and readable type
  • Enough frontend to ship a working prototype without a handoff queue

Tools you should already use

FigmaHTML / CSSSwiftUI or UIKit basicsSF SymbolswatchOS glance layoutsGitContrast checkersPrototyping in-browser
Apply · [email protected]

Connectors engineer

Full-timeRemoteOAuth

Mail, calendar, docs, chat. Scoped OAuth, local token store, revoke per app. Off means the tool is never offered to the model. Hosted extras stay metered.

You will

  • Ship Gmail, Calendar, Slack, Drive, GitHub and similar as toggles in the chat
  • Keep tokens in a local key store and make revoke a one-tap action
  • Generate the tool list from capability switches so a disabled app is invisible to the model
  • Meter hosted connector calls without taxing the free local loop

Skills we need

  • OAuth 2.0 / OIDC, refresh rotation, and scoped consent screens
  • REST and webhook design that fails closed
  • Mapping third-party APIs onto a small, strict tool schema
  • Privacy instinct: the model only sees what the user switched on

Tools you should already use

OAuth 2.0OIDCGmail APIGoogle CalendarSlack APIMicrosoft GraphGitHub AppsKeychain / DPAPIngrokPostman or Bruno
Apply · [email protected]

Apple platforms engineer

Full-timeRemoteApple

Same loop on macOS, iOS, and watchOS. Folder manager on Mac. Chat, inbound number, and approvals on iPhone. Glance a task and confirm a write from the wrist.

You will

  • Ship the Apple binary: Mac folder manager, iOS chat, watchOS glance and confirm
  • Keep first-class performance on Apple Silicon; Intel Macs use the same install
  • Wire local model servers on the box or LAN without dragging files off-device
  • Handle background delivery for inbound SMS / WhatsApp when hosted extras are on

Skills we need

  • Native Apple apps, not a thin web wrapper that dies in the background
  • Sandbox, entitlements, Keychain, and App Groups across Mac, iPhone, Watch
  • File coordination against one workspace root
  • Enough networking to talk to localhost model servers cleanly

Tools you should already use

SwiftSwiftUIUIKitWatchKitXcodeKeychainApp SandboxCloudKit optionalInstrumentsTestFlight
Apply · [email protected]

Evaluation engineer

Full-timeRemoteQuality

Folder evals that finish a real task, not a demo. Kill tricks that only look good on this site. Hold public beta until private-trial bugs are gone.

You will

  • Write eval suites against real workspaces: edit a repo, summarise a folder, draft then wait to send
  • Score plan quality, tool choice, and whether the gate fired at the right time
  • Compare local weights vs hosted fallback without leaking files
  • Block a release when a private-trial bug is still open

Skills we need

  • Eval harnesses, fixtures, and golden transcripts you can diff
  • Statistical sense: a single pretty run is not a result
  • Comfort reading model traces and naming the failure mode
  • Product judgement for “did the task actually finish”

Tools you should already use

PythonpytestJSONL fixturesOllamaOpenAI-compatible APIsGitR or pandasCI (GitHub Actions)markdown reports
Apply · [email protected]

How to apply

Email [email protected]. Put the role in the subject, for example Application — Runtime engineer. One thread. No form, no recruiter portal.

  • A short note on what you have shipped that is closest to this product
  • Links: GitHub, site, App Store, paper, or a folder you can talk through
  • The tools from the list above you actually use week to week, not a keyword dump
  • Where you work from, and the hours that overlap NZ / AU / UK / US
  • General applications are fine if no role fits. Same inbox, subject: General application
Send application · [email protected]